On the performance analysis of short LDPC codes

摘要

This paper presents the research on LDPC codes performance depending on several Tanner graph properties. We have simulated 2250 (128,256) different LDPC codes in FPGA and explored their waterfall region as a function of the graph characteristics. These characteristics are different graph centralization measures and the second smallest Laplacian eigenvalue. We have received numerical values that distinguish LDPC codes with a good decoding threshold (less or equal 3.5 dB) from others. In addition, we developed a polynomial regression model that accurately predicts the LDPC decoding threshold with a coefficient of determination equal to 0.87 and RMSE equal 0.16 dB.